Prediction Table History Selection for Lower Lookup Power
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data processing apparatuses waste significant power by treating all prediction tables as active during program execution, despite some tables not contributing to predictions, leading to unnecessary toggling and power consumption in lookup information generation circuitry.
Innovation Solution
Implement input history selection circuitry to prevent a given portion of input history information from differing between predictions, thereby reducing power consumption by treating certain prediction tables as inactive, especially those using longer history information, and using power gating or snapshot circuitry to maintain consistent input values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If all prediction tables are treated as active during program execution, then prediction accuracy is maintained, but power consumption increases due to unnecessary logic toggling in lookup information generation circuitry
Solution Approach 1:
The patent segments prediction tables into active and inactive subsets based on their contribution to predictions. Lookup information generation circuitry is further segmented to independently generate lookup information for active tables (using full history information) and inactive tables (using reduced history information), preventing unnecessary logic toggling while maintaining prediction accuracy through selective activation.
Solution Approach 2:
The patent implements dynamic adjustment of prediction table activity status based on runtime conditions. Input history selection circuitry dynamically determines which prediction tables should be active or inactive for each prediction, allowing the system to adapt the configuration of lookup information generation circuitry to minimize power consumption while maintaining required prediction accuracy.
2Power
If lookup information generation circuitry processes all prediction tables using full history information, then prediction quality is maintained, but dynamic power consumption increases due to unnecessary logic toggling
Solution Approach 1:
The patent applies local quality by treating different prediction tables differently based on their specific characteristics and contribution to predictions. Active prediction tables receive full history information processing for high prediction quality, while inactive tables receive reduced history information processing for lower power consumption, optimizing the balance between quality and energy efficiency locally for each table.
Solution Approach 2:
The patent applies partial action by processing only the necessary portion of history information for inactive prediction tables rather than full history information. This reduces the computational effort and logic toggling in lookup information generation circuitry for inactive tables while maintaining sufficient prediction quality through the partial history processing of active tables.
Data Source
AI summary
An apparatus comprises a predictor to make a prediction based on a plurality of prediction tables. The plurality of prediction tables are looked up using table lookup information generated based on different lengths of input history information representing a path through program execution. The apparatus comprises circuitry to prevent a given portion of the input history information from differing with respect to a corresponding portion of the input history information used to make a preceding prediction, where the given portion is a portion which is not used to generate table lookup information for an active subset of the prediction tables.